OBJECTIVE: To advance radiological techniques for detection and diagnosis of cancer pathology, with potentially reduced exposure of the patient to radiation, by developing unconventional measurement techniques and associated reconstruction algorithms for computer assisted tomography (CAT). APPROACH: Methods of X-ray projection data collection will be investigated which involve modified forms of the conventional scanning modes. Projections are usually measured over their complete linear extent, i.e., the measurement field of view includes the entire cross section under examination. However, only a specific region of the cross-sectional image needs to be reconstructed accurately for the detection and diagnosis of some types of early cancer lesions. Hence the X-ray dose to the patient would be reduced, compared to that of conventional CAT, if projection data were accurately measured only for X-rays traversing the cross-sectional region of interest. Parts of the projections which are outside the limited field of view might be measured at low spatial resolution, or not measured at all. The mathematical basis of image reconstruction from such data will be formulated, and practical algorithms will be implemented and evaluated. The performance of various algorithms will be compared using computer-generated projections of realistic simulated phantoms. Measured projection data will be used to determine how these algorithms should be applied in clinical practice.